Generally, in a large software system such as software as a service (SaaS) and a large-scale distribution system, mass log messages may be generated each day for different purposes such as system monitoring, user behavior recording, and troubleshooting, etc. Typically, such log data may be generated at a very high throughput which may usually reach, for example, 10 GB per hour, and such log traffic may occupy a considerable I/O capacity and very likely become a bottleneck for system performance.